linux/drivers/cpuidle/cpuidle-pseries.c

287 lines
6.2 KiB
C
Raw Normal View History

/*
* cpuidle-pseries - idle state cpuidle driver.
* Adapted from drivers/idle/intel_idle.c and
* drivers/acpi/processor_idle.c
*
*/
#include <linux/kernel.h>
#include <linux/module.h>
#include <linux/init.h>
#include <linux/moduleparam.h>
#include <linux/cpuidle.h>
#include <linux/cpu.h>
#include <linux/notifier.h>
#include <asm/paca.h>
#include <asm/reg.h>
#include <asm/machdep.h>
#include <asm/firmware.h>
#include <asm/runlatch.h>
#include <asm/plpar_wrappers.h>
struct cpuidle_driver pseries_idle_driver = {
.name = "pseries_idle",
.owner = THIS_MODULE,
};
static int max_idle_state __read_mostly;
static struct cpuidle_state *cpuidle_state_table __read_mostly;
static u64 snooze_timeout __read_mostly;
static bool snooze_timeout_en __read_mostly;
static inline void idle_loop_prolog(unsigned long *in_purr)
{
sched/idle, PPC: Remove redundant cpuidle_idle_call() The core idle loop now takes care of it. However a few things need checking: - Invocation of cpuidle_idle_call() in pseries_lpar_idle() happened through arch_cpu_idle() and was therefore always preceded by a call to ppc64_runlatch_off(). To preserve this property now that cpuidle_idle_call() is invoked directly from core code, a call to ppc64_runlatch_off() has been added to idle_loop_prolog() in platforms/pseries/processor_idle.c. - Similarly, cpuidle_idle_call() was followed by ppc64_runlatch_off() so a call to the later has been added to idle_loop_epilog(). - And since arch_cpu_idle() always made sure to re-enable IRQs if they were not enabled, this is now done in idle_loop_epilog() as well. The above was made in order to keep the execution flow close to the original. I don't know if that was strictly necessary. Someone well aquainted with the platform details might find some room for possible optimizations. Signed-off-by: Nicolas Pitre <nico@linaro.org> Reviewed-by: Preeti U Murthy <preeti@linux.vnet.ibm.com> Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net> Cc: Daniel Lezcano <daniel.lezcano@linaro.org> Cc: linux-arm-kernel@lists.infradead.org Cc: linuxppc-dev@lists.ozlabs.org Cc: linux-sh@vger.kernel.org Cc: linux-pm@vger.kernel.org Cc: Russell King <linux@arm.linux.org.uk> Cc: linaro-kernel@lists.linaro.org Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mundt <lethal@linux-sh.org> Signed-off-by: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/n/tip-47o4m03citrfg9y1vxic5asb@git.kernel.org Signed-off-by: Ingo Molnar <mingo@kernel.org>
2014-01-30 01:45:10 +08:00
ppc64_runlatch_off();
*in_purr = mfspr(SPRN_PURR);
/*
* Indicate to the HV that we are idle. Now would be
* a good time to find other work to dispatch.
*/
get_lppaca()->idle = 1;
}
static inline void idle_loop_epilog(unsigned long in_purr)
{
u64 wait_cycles;
wait_cycles = be64_to_cpu(get_lppaca()->wait_state_cycles);
wait_cycles += mfspr(SPRN_PURR) - in_purr;
get_lppaca()->wait_state_cycles = cpu_to_be64(wait_cycles);
get_lppaca()->idle = 0;
sched/idle, PPC: Remove redundant cpuidle_idle_call() The core idle loop now takes care of it. However a few things need checking: - Invocation of cpuidle_idle_call() in pseries_lpar_idle() happened through arch_cpu_idle() and was therefore always preceded by a call to ppc64_runlatch_off(). To preserve this property now that cpuidle_idle_call() is invoked directly from core code, a call to ppc64_runlatch_off() has been added to idle_loop_prolog() in platforms/pseries/processor_idle.c. - Similarly, cpuidle_idle_call() was followed by ppc64_runlatch_off() so a call to the later has been added to idle_loop_epilog(). - And since arch_cpu_idle() always made sure to re-enable IRQs if they were not enabled, this is now done in idle_loop_epilog() as well. The above was made in order to keep the execution flow close to the original. I don't know if that was strictly necessary. Someone well aquainted with the platform details might find some room for possible optimizations. Signed-off-by: Nicolas Pitre <nico@linaro.org> Reviewed-by: Preeti U Murthy <preeti@linux.vnet.ibm.com> Cc: "Rafael J. Wysocki" <rjw@rjwysocki.net> Cc: Daniel Lezcano <daniel.lezcano@linaro.org> Cc: linux-arm-kernel@lists.infradead.org Cc: linuxppc-dev@lists.ozlabs.org Cc: linux-sh@vger.kernel.org Cc: linux-pm@vger.kernel.org Cc: Russell King <linux@arm.linux.org.uk> Cc: linaro-kernel@lists.linaro.org Cc: Benjamin Herrenschmidt <benh@kernel.crashing.org> Cc: Paul Mundt <lethal@linux-sh.org> Signed-off-by: Peter Zijlstra <peterz@infradead.org> Link: http://lkml.kernel.org/n/tip-47o4m03citrfg9y1vxic5asb@git.kernel.org Signed-off-by: Ingo Molnar <mingo@kernel.org>
2014-01-30 01:45:10 +08:00
if (irqs_disabled())
local_irq_enable();
ppc64_runlatch_on();
}
static int snooze_loop(struct cpuidle_device *dev,
struct cpuidle_driver *drv,
int index)
{
unsigned long in_purr;
u64 snooze_exit_time;
set_thread_flag(TIF_POLLING_NRFLAG);
idle_loop_prolog(&in_purr);
cpuidle/powerpc: Fix snooze state problem in the cpuidle design on pseries. Earlier without cpuidle framework on pseries, the native arch idle routine comprised of both snooze and nap states. smt_snooze_delay variable was used to delay the idle process entry to deeper idle state like nap. With the coming of cpuidle, this arch specific idle was replaced by two different idle routines, one for supporting snooze and other for nap. This enabled addition of more low level idle states on pseries in the future. On adopting the generic cpuidle framework for POWER systems, the decision of which idle state to choose from, given a predicted idle time is taken by the menu governor based on target_residency and exit_latency of the idle states. target_residency is the minimum time to be resident in that idle state. Exit_latency is time taken to exit out of idle state. Deeper the idle state, both the target residency and exit latency would be higher. In the current design, smt_snooze_delay is used as target_residency for the snooze state which is incorrect, as it is not the minimum but the maximum duration to be in snooze state. This would result in the governor in taking bad decision, as presently target_residency of nap < target_residency of snooze inspite of nap being deeper idle state. This patch aims to fix this problem by replacing the smt_snooze_delay loop in snooze state, with the need_resched() as the governor is aware of entry and exit of various idle transitions based on which next idle time prediction. The governor is intelligent enough to determine the idle state the needs to be transitioned to and maintains a whole of heuristics including io load, previous idle states predictions etc for the same, based on which idle state entry decision is taken. With this fix, of setting target_residency of snooze to 0 nap to smt_snooze_delay if the predicted idle time is less than smt_snooze_delay (target_residency of nap) value governor would pick snooze state, else nap. This adhers to the previous native idle design. Signed-off-by: Deepthi Dharwar <deepthi@linux.vnet.ibm.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-10-04 02:42:26 +08:00
local_irq_enable();
snooze_exit_time = get_tb() + snooze_timeout;
while (!need_resched()) {
cpuidle/powerpc: Fix snooze state problem in the cpuidle design on pseries. Earlier without cpuidle framework on pseries, the native arch idle routine comprised of both snooze and nap states. smt_snooze_delay variable was used to delay the idle process entry to deeper idle state like nap. With the coming of cpuidle, this arch specific idle was replaced by two different idle routines, one for supporting snooze and other for nap. This enabled addition of more low level idle states on pseries in the future. On adopting the generic cpuidle framework for POWER systems, the decision of which idle state to choose from, given a predicted idle time is taken by the menu governor based on target_residency and exit_latency of the idle states. target_residency is the minimum time to be resident in that idle state. Exit_latency is time taken to exit out of idle state. Deeper the idle state, both the target residency and exit latency would be higher. In the current design, smt_snooze_delay is used as target_residency for the snooze state which is incorrect, as it is not the minimum but the maximum duration to be in snooze state. This would result in the governor in taking bad decision, as presently target_residency of nap < target_residency of snooze inspite of nap being deeper idle state. This patch aims to fix this problem by replacing the smt_snooze_delay loop in snooze state, with the need_resched() as the governor is aware of entry and exit of various idle transitions based on which next idle time prediction. The governor is intelligent enough to determine the idle state the needs to be transitioned to and maintains a whole of heuristics including io load, previous idle states predictions etc for the same, based on which idle state entry decision is taken. With this fix, of setting target_residency of snooze to 0 nap to smt_snooze_delay if the predicted idle time is less than smt_snooze_delay (target_residency of nap) value governor would pick snooze state, else nap. This adhers to the previous native idle design. Signed-off-by: Deepthi Dharwar <deepthi@linux.vnet.ibm.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-10-04 02:42:26 +08:00
HMT_low();
HMT_very_low();
if (likely(snooze_timeout_en) && get_tb() > snooze_exit_time) {
/*
* Task has not woken up but we are exiting the polling
* loop anyway. Require a barrier after polling is
* cleared to order subsequent test of need_resched().
*/
clear_thread_flag(TIF_POLLING_NRFLAG);
smp_mb();
break;
}
}
HMT_medium();
cpuidle/powerpc: Fix snooze state problem in the cpuidle design on pseries. Earlier without cpuidle framework on pseries, the native arch idle routine comprised of both snooze and nap states. smt_snooze_delay variable was used to delay the idle process entry to deeper idle state like nap. With the coming of cpuidle, this arch specific idle was replaced by two different idle routines, one for supporting snooze and other for nap. This enabled addition of more low level idle states on pseries in the future. On adopting the generic cpuidle framework for POWER systems, the decision of which idle state to choose from, given a predicted idle time is taken by the menu governor based on target_residency and exit_latency of the idle states. target_residency is the minimum time to be resident in that idle state. Exit_latency is time taken to exit out of idle state. Deeper the idle state, both the target residency and exit latency would be higher. In the current design, smt_snooze_delay is used as target_residency for the snooze state which is incorrect, as it is not the minimum but the maximum duration to be in snooze state. This would result in the governor in taking bad decision, as presently target_residency of nap < target_residency of snooze inspite of nap being deeper idle state. This patch aims to fix this problem by replacing the smt_snooze_delay loop in snooze state, with the need_resched() as the governor is aware of entry and exit of various idle transitions based on which next idle time prediction. The governor is intelligent enough to determine the idle state the needs to be transitioned to and maintains a whole of heuristics including io load, previous idle states predictions etc for the same, based on which idle state entry decision is taken. With this fix, of setting target_residency of snooze to 0 nap to smt_snooze_delay if the predicted idle time is less than smt_snooze_delay (target_residency of nap) value governor would pick snooze state, else nap. This adhers to the previous native idle design. Signed-off-by: Deepthi Dharwar <deepthi@linux.vnet.ibm.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-10-04 02:42:26 +08:00
clear_thread_flag(TIF_POLLING_NRFLAG);
idle_loop_epilog(in_purr);
return index;
}
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 15:27:59 +08:00
static void check_and_cede_processor(void)
{
/*
* Ensure our interrupt state is properly tracked,
* also checks if no interrupt has occurred while we
* were soft-disabled
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 15:27:59 +08:00
*/
if (prep_irq_for_idle()) {
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 15:27:59 +08:00
cede_processor();
#ifdef CONFIG_TRACE_IRQFLAGS
/* Ensure that H_CEDE returns with IRQs on */
if (WARN_ON(!(mfmsr() & MSR_EE)))
__hard_irq_enable();
#endif
}
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 15:27:59 +08:00
}
static int dedicated_cede_loop(struct cpuidle_device *dev,
struct cpuidle_driver *drv,
int index)
{
unsigned long in_purr;
idle_loop_prolog(&in_purr);
get_lppaca()->donate_dedicated_cpu = 1;
HMT_medium();
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 15:27:59 +08:00
check_and_cede_processor();
get_lppaca()->donate_dedicated_cpu = 0;
idle_loop_epilog(in_purr);
return index;
}
static int shared_cede_loop(struct cpuidle_device *dev,
struct cpuidle_driver *drv,
int index)
{
unsigned long in_purr;
idle_loop_prolog(&in_purr);
/*
* Yield the processor to the hypervisor. We return if
* an external interrupt occurs (which are driven prior
* to returning here) or if a prod occurs from another
* processor. When returning here, external interrupts
* are enabled.
*/
powerpc: Rework lazy-interrupt handling The current implementation of lazy interrupts handling has some issues that this tries to address. We don't do the various workarounds we need to do when re-enabling interrupts in some cases such as when returning from an interrupt and thus we may still lose or get delayed decrementer or doorbell interrupts. The current scheme also makes it much harder to handle the external "edge" interrupts provided by some BookE processors when using the EPR facility (External Proxy) and the Freescale Hypervisor. Additionally, we tend to keep interrupts hard disabled in a number of cases, such as decrementer interrupts, external interrupts, or when a masked decrementer interrupt is pending. This is sub-optimal. This is an attempt at fixing it all in one go by reworking the way we do the lazy interrupt disabling from the ground up. The base idea is to replace the "hard_enabled" field with a "irq_happened" field in which we store a bit mask of what interrupt occurred while soft-disabled. When re-enabling, either via arch_local_irq_restore() or when returning from an interrupt, we can now decide what to do by testing bits in that field. We then implement replaying of the missed interrupts either by re-using the existing exception frame (in exception exit case) or via the creation of a new one from an assembly trampoline (in the arch_local_irq_enable case). This removes the need to play with the decrementer to try to create fake interrupts, among others. In addition, this adds a few refinements: - We no longer hard disable decrementer interrupts that occur while soft-disabled. We now simply bump the decrementer back to max (on BookS) or leave it stopped (on BookE) and continue with hard interrupts enabled, which means that we'll potentially get better sample quality from performance monitor interrupts. - Timer, decrementer and doorbell interrupts now hard-enable shortly after removing the source of the interrupt, which means they no longer run entirely hard disabled. Again, this will improve perf sample quality. - On Book3E 64-bit, we now make the performance monitor interrupt act as an NMI like Book3S (the necessary C code for that to work appear to already be present in the FSL perf code, notably calling nmi_enter instead of irq_enter). (This also fixes a bug where BookE perfmon interrupts could clobber r14 ... oops) - We could make "masked" decrementer interrupts act as NMIs when doing timer-based perf sampling to improve the sample quality. Signed-off-by-yet: Benjamin Herrenschmidt <benh@kernel.crashing.org> --- v2: - Add hard-enable to decrementer, timer and doorbells - Fix CR clobber in masked irq handling on BookE - Make embedded perf interrupt act as an NMI - Add a PACA_HAPPENED_EE_EDGE for use by FSL if they want to retrigger an interrupt without preventing hard-enable v3: - Fix or vs. ori bug on Book3E - Fix enabling of interrupts for some exceptions on Book3E v4: - Fix resend of doorbells on return from interrupt on Book3E v5: - Rebased on top of my latest series, which involves some significant rework of some aspects of the patch. v6: - 32-bit compile fix - more compile fixes with various .config combos - factor out the asm code to soft-disable interrupts - remove the C wrapper around preempt_schedule_irq v7: - Fix a bug with hard irq state tracking on native power7
2012-03-06 15:27:59 +08:00
check_and_cede_processor();
idle_loop_epilog(in_purr);
return index;
}
/*
* States for dedicated partition case.
*/
static struct cpuidle_state dedicated_states[] = {
{ /* Snooze */
.name = "snooze",
.desc = "snooze",
.exit_latency = 0,
.target_residency = 0,
.enter = &snooze_loop },
{ /* CEDE */
.name = "CEDE",
.desc = "CEDE",
cpuidle/powerpc: Fix snooze state problem in the cpuidle design on pseries. Earlier without cpuidle framework on pseries, the native arch idle routine comprised of both snooze and nap states. smt_snooze_delay variable was used to delay the idle process entry to deeper idle state like nap. With the coming of cpuidle, this arch specific idle was replaced by two different idle routines, one for supporting snooze and other for nap. This enabled addition of more low level idle states on pseries in the future. On adopting the generic cpuidle framework for POWER systems, the decision of which idle state to choose from, given a predicted idle time is taken by the menu governor based on target_residency and exit_latency of the idle states. target_residency is the minimum time to be resident in that idle state. Exit_latency is time taken to exit out of idle state. Deeper the idle state, both the target residency and exit latency would be higher. In the current design, smt_snooze_delay is used as target_residency for the snooze state which is incorrect, as it is not the minimum but the maximum duration to be in snooze state. This would result in the governor in taking bad decision, as presently target_residency of nap < target_residency of snooze inspite of nap being deeper idle state. This patch aims to fix this problem by replacing the smt_snooze_delay loop in snooze state, with the need_resched() as the governor is aware of entry and exit of various idle transitions based on which next idle time prediction. The governor is intelligent enough to determine the idle state the needs to be transitioned to and maintains a whole of heuristics including io load, previous idle states predictions etc for the same, based on which idle state entry decision is taken. With this fix, of setting target_residency of snooze to 0 nap to smt_snooze_delay if the predicted idle time is less than smt_snooze_delay (target_residency of nap) value governor would pick snooze state, else nap. This adhers to the previous native idle design. Signed-off-by: Deepthi Dharwar <deepthi@linux.vnet.ibm.com> Signed-off-by: Benjamin Herrenschmidt <benh@kernel.crashing.org>
2012-10-04 02:42:26 +08:00
.exit_latency = 10,
.target_residency = 100,
.enter = &dedicated_cede_loop },
};
/*
* States for shared partition case.
*/
static struct cpuidle_state shared_states[] = {
{ /* Shared Cede */
.name = "Shared Cede",
.desc = "Shared Cede",
.exit_latency = 0,
.target_residency = 0,
.enter = &shared_cede_loop },
};
static int pseries_cpuidle_cpu_online(unsigned int cpu)
{
struct cpuidle_device *dev = per_cpu(cpuidle_devices, cpu);
if (dev && cpuidle_get_driver()) {
cpuidle_pause_and_lock();
cpuidle_enable_device(dev);
cpuidle_resume_and_unlock();
}
return 0;
}
static int pseries_cpuidle_cpu_dead(unsigned int cpu)
{
struct cpuidle_device *dev = per_cpu(cpuidle_devices, cpu);
if (dev && cpuidle_get_driver()) {
cpuidle_pause_and_lock();
cpuidle_disable_device(dev);
cpuidle_resume_and_unlock();
}
return 0;
}
/*
* pseries_cpuidle_driver_init()
*/
static int pseries_cpuidle_driver_init(void)
{
int idle_state;
struct cpuidle_driver *drv = &pseries_idle_driver;
drv->state_count = 0;
for (idle_state = 0; idle_state < max_idle_state; ++idle_state) {
/* Is the state not enabled? */
if (cpuidle_state_table[idle_state].enter == NULL)
continue;
drv->states[drv->state_count] = /* structure copy */
cpuidle_state_table[idle_state];
drv->state_count += 1;
}
return 0;
}
/*
* pseries_idle_probe()
* Choose state table for shared versus dedicated partition
*/
static int pseries_idle_probe(void)
{
if (cpuidle_disable != IDLE_NO_OVERRIDE)
return -ENODEV;
if (firmware_has_feature(FW_FEATURE_SPLPAR)) {
if (lppaca_shared_proc(get_lppaca())) {
cpuidle_state_table = shared_states;
max_idle_state = ARRAY_SIZE(shared_states);
} else {
cpuidle_state_table = dedicated_states;
max_idle_state = ARRAY_SIZE(dedicated_states);
}
} else
return -ENODEV;
if (max_idle_state > 1) {
snooze_timeout_en = true;
snooze_timeout = cpuidle_state_table[1].target_residency *
tb_ticks_per_usec;
}
return 0;
}
static int __init pseries_processor_idle_init(void)
{
int retval;
retval = pseries_idle_probe();
if (retval)
return retval;
pseries_cpuidle_driver_init();
retval = cpuidle_register(&pseries_idle_driver, NULL);
if (retval) {
printk(KERN_DEBUG "Registration of pseries driver failed.\n");
return retval;
}
retval = cpuhp_setup_state_nocalls(CPUHP_AP_ONLINE_DYN,
"cpuidle/pseries:online",
pseries_cpuidle_cpu_online, NULL);
WARN_ON(retval < 0);
retval = cpuhp_setup_state_nocalls(CPUHP_CPUIDLE_DEAD,
"cpuidle/pseries:DEAD", NULL,
pseries_cpuidle_cpu_dead);
WARN_ON(retval < 0);
printk(KERN_DEBUG "pseries_idle_driver registered\n");
return 0;
}
device_initcall(pseries_processor_idle_init);